This git is related to a study on reward function for reinforcement learning for a micro-grid.

The last environment used. it inherits from the gym Env class. Our environment contains information stored in a dictionary. It contains everything that is in _init_dict:

• "flow_H2"
• "flow_lithium"
• "lack_energy"
• "waste_energy"
• "soc"
• "buy_energy"
• "sell_energy"

As you can see, all parameter names used must be in lower case and with a unique name if you want to modify the environment. You will also have to normalize your values in _normalize_value method.

To add reward functions to your environment, you can use

add_reward(key, fn, coeff=1.)

The data is stored and retrievable from

get_data()

This folder contains the useful files for the genetic algorithm and the Map-Elites.

Contains the functions for the genetic algorithm.

Contains the funtions for Map-Elites.

This is the place where you can run reward function tests. The places you have to modify are :

• function in creat_lfn (Not _set_fn and _get_mean but below)

  • exemple (lfn and lcut is created before)

  isNeg = True
  min_val= -1
  max_val = 0
  
  # function name must start with a capital letter and be unique
  lname.append("Name_exemple")  
  fn, cut = _set_fn(lambda x: -x["buy_energy"], lres_reset, isNeg, min_val, max_val, nb_cut)  
  #or 
  fn, cut = _set_fn(lambda x, f=my_fn: f(x), lres_reset, isNeg, min_val, max_val, nb_cut)
  lfn.append(fn)  
  lcut.append(cut)
  

• "values has initialized" part in main loop